Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Metadata Integration for Spam Reviews Detection on Vietnamese E-commerce Websites

Co Van Dinh, Son T. Luu

TL;DR

This work introduces ViSpamReviews v2, a metadata-rich dataset for spam review detection on Vietnamese e-commerce, and a two-phase metadata integration approach that fuses textual embeddings from SBERT/SPhoBERT with categorical product signals. By evaluating both DNN and BERT-based models, the study shows that product descriptions primarily boost transformer-based models while product category benefits DNNs, with the best overall results achieved by PhoBERT enhanced with SPhoBert-derived product descriptions (Task 1: 90.11% accuracy, macro F1 87.22%; Task 2: 88.06% accuracy, macro F1 73.49%). The findings highlight the value of metadata in improving spam detection performance and offer practical guidance on when to leverage description versus category signals across model families. The work also discusses limitations and suggests combining rule-based approaches with learned models to further enhance detection, particularly for SPAM-2 and SPAM-3 types.

Abstract

The problem of detecting spam reviews (opinions) has received significant attention in recent years, especially with the rapid development of e-commerce. Spam reviews are often classified based on comment content, but in some cases, it is insufficient for models to accurately determine the review label. In this work, we introduce the ViSpamReviews v2 dataset, which includes metadata of reviews with the objective of integrating supplementary attributes for spam review classification. We propose a novel approach to simultaneously integrate both textual and categorical attributes into the classification model. In our experiments, the product category proved effective when combined with deep neural network (DNN) models, while text features performed well on both DNN models and the model achieved state-of-the-art performance in the problem of detecting spam reviews on Vietnamese e-commerce websites, namely PhoBERT. Specifically, the PhoBERT model achieves the highest accuracy when combined with product description features generated from the SPhoBert model, which is the combination of PhoBERT and SentenceBERT. Using the macro-averaged F1 score, the task of classifying spam reviews achieved 87.22% (an increase of 1.64% compared to the baseline), while the task of identifying the type of spam reviews achieved an accuracy of 73.49% (an increase of 1.93% compared to the baseline).

Metadata Integration for Spam Reviews Detection on Vietnamese E-commerce Websites

TL;DR

This work introduces ViSpamReviews v2, a metadata-rich dataset for spam review detection on Vietnamese e-commerce, and a two-phase metadata integration approach that fuses textual embeddings from SBERT/SPhoBERT with categorical product signals. By evaluating both DNN and BERT-based models, the study shows that product descriptions primarily boost transformer-based models while product category benefits DNNs, with the best overall results achieved by PhoBERT enhanced with SPhoBert-derived product descriptions (Task 1: 90.11% accuracy, macro F1 87.22%; Task 2: 88.06% accuracy, macro F1 73.49%). The findings highlight the value of metadata in improving spam detection performance and offer practical guidance on when to leverage description versus category signals across model families. The work also discusses limitations and suggests combining rule-based approaches with learned models to further enhance detection, particularly for SPAM-2 and SPAM-3 types.

Abstract

The problem of detecting spam reviews (opinions) has received significant attention in recent years, especially with the rapid development of e-commerce. Spam reviews are often classified based on comment content, but in some cases, it is insufficient for models to accurately determine the review label. In this work, we introduce the ViSpamReviews v2 dataset, which includes metadata of reviews with the objective of integrating supplementary attributes for spam review classification. We propose a novel approach to simultaneously integrate both textual and categorical attributes into the classification model. In our experiments, the product category proved effective when combined with deep neural network (DNN) models, while text features performed well on both DNN models and the model achieved state-of-the-art performance in the problem of detecting spam reviews on Vietnamese e-commerce websites, namely PhoBERT. Specifically, the PhoBERT model achieves the highest accuracy when combined with product description features generated from the SPhoBert model, which is the combination of PhoBERT and SentenceBERT. Using the macro-averaged F1 score, the task of classifying spam reviews achieved 87.22% (an increase of 1.64% compared to the baseline), while the task of identifying the type of spam reviews achieved an accuracy of 73.49% (an increase of 1.93% compared to the baseline).
Paper Structure (17 sections, 13 equations, 7 figures, 9 tables)

This paper contains 17 sections, 13 equations, 7 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. The Dataset
  5. Metadata Analysis
  6. Metadata Integration
  7. Product Description
  8. Product Category
  9. Results and Discussion
  10. Data Preprocessing
  11. Experiment Settings
  12. Evaluation Metrics
  13. Task 1
  14. Task 2
  15. Empirical Results
  16. ...and 2 more sections

Figures (7)

  • Figure 1: The metadata is collected from Shopee.
  • Figure 2: Distribution of labels for reviews by product category.
  • Figure 3: Distribution of product name and product description lengths (calculated based on word count) on the training set.
  • Figure 4: An overview of the proposed method for integrating metadata in spam reviews detection.
  • Figure 5: Data preprocessing pipeline.
  • ...and 2 more figures